T H E S I N C E R E V O T E

Transcription

1 THE SINCERE VOTE

2 Printed by Febodruk b.v., Enschede Copyright 2004 by Martin Rosema. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the written permission of the author.

3 THE SINCERE VOTE A Psychological Study of Voting Proefschrift ter verkrijging van de graad van Doctor aan de Universiteit Leiden, op gezag van de Rector Magnificus Dr. D. D. Breimer, hoogleraar in de faculteit der Wiskunde en Natuurwetenschappen en die der Geneeskunde, volgens besluit van het College voor Promoties te verdedigen op dinsdag 23 november 2004 klokke uur door Martin Rosema geboren te Hengelo (Ov.) in 1970

4 PROMOTIECOMMISSIE promotor: Prof. G. A. Irwin Ph.D. copromotor: Dr. J. J. M. van Holsteyn referent: Prof. dr. R. B. Andeweg overige leden: Prof. dr. C. M. J. G. Maes Prof. dr. J. J. A. Thomassen

5 PREFACE Elections have long fascinated me. This interest was one of the driving forces behind my decision in the early 1990s to study political science in Leiden. In those years, I soon came to realise that if one wants to understand the choices of voters, some knowledge of psychology might be helpful. So I followed a couple of introductory courses in that field. When I had to choose a subject for my master s thesis, I ultimately decided to focus on what interested me most and apply some ideas from psychology to the electoral context. When I had finished the thesis, my conclusion was that the strategy adopted had been fruitful. Moreover, in as far as I did not exactly know why people voted in a particular way, at least I had some ideas about how to find out. I wrote down my ideas in a research proposal, on the basis of which I was awarded the position of Ph.D. candidate in the department of Political Science at the University of Leiden. It took a little longer than I anticipated, but the final result now lies in front of you. Of the many things that could additionally be said, there is one peculiarity that seems nice to highlight. During my doctoraal (three-year program leading to a master s degree) in political science at the University of Leiden, there were two courses for which I received the lowest passing grade. The first course was on research methods, and the second was on political psychology. Strikingly, both topics feature quite prominently in this thesis. Perhaps this makes you wonder whether my dissertation would have been much better, had it been on another subject. Or perhaps you wonder how much better a dissertation would have been written about this subject, had someone else been given the opportunity to do so. These are interesting questions, although we might never know the answer. This leads me to the people whose contribution I wish to acknowledge. First, I want to express my gratitude to the department of Political Science at the University of Twente, which generously provided me with the opportunity to finish the disser- v

6 tation in a stimulating and friendly environment. Second, I am grateful to the colleagues who commented on draft versions of chapters or related papers, which I presented at several conferences. Third, I would like to thank the person who was willing to use his artistic talents, which are not easily underestimated, for designing the cover of this thesis: Martin van Leeuwen. Finally, I would not have been able to write this thesis if my private life would not have been such a happy one. Family and friends contributed to this, but the single person largely responsible is the one who accompanied me in life for all those years. Enschede, October 2004 Martin Rosema vi

7 CONTENTS IN BRIEF Preface Contents in detail List of figures and tables v ix xv 1. Introduction 3 2. Psychology in voting theory 9 3. Attitude-behaviour models and voting The sincere vote model Vote choice heuristics Three models to explain party evaluations Empirical test of the sincere vote model The non-sincere vote Explaining party evaluations: a traditional approach A psychological theory of voting 201 Appendices 221 Notes 259 Bibliography 289 Author index 305 Nederlandse samenvatting (Dutch summary) 311 Curriculum vitae 329 vii

8

9 CONTENTS IN DETAIL PART I: INTRODUCTION AND BACKGROUND 1. Introduction 3 The research question 4 A psychological approach 5 The Dutch case 7 Design of the book 8 2. Psychology in Voting Theory 9 On electoral research 9 development of modern electoral research the force field framework the spatial framework A sociological perspective: the Columbia studies 13 social characteristics and political predisposition cleavages in western Europe social characteristics and the psychological perspective A psychological perspective: the Michigan studies 18 party identification, candidate orientation, and issue orientation funnel of causality and partisan attitudes candidate evaluations as direct determinants of vote choice the Michigan studies and the psychological perspective party identification in the Netherlands ix

10 An economic perspective: Downs and spatial models 27 an economic theory of democracy spatial models of ideological voting in the Netherlands Downs theory and the psychological perspective Theories of issue voting 34 the issue-oriented, responsible voter proximity model of issue voting directional theory of issue voting theory of issue ownership theories of issue voting and the psychological perspective The psephological paradigm 39 the assumption of voting as a two-decision process the assumption of a single object of voting the assumption of parties as single, unitary actors the assumption of causal homogeneity in vote choice the assumption of a sincere vote the assumption of memory-based candidate and party evaluations the assumption of cognitive and semantic memory the assumption of homogeneity in bases of evaluation (across parties) the assumption of constructable evaluations the assumption of researcher foreknowledge the psychological process as a black box 3. Attitude-Behaviour Models and Voting 47 The concept of an attitude 47 conceptual definitions of attitudes operational definitions of attitudes in voting research Attitudes and behaviour 50 the attitude-behaviour relationship theory of reasoned action mode model and composite model Attitudes and voting behaviour 55 the kind of attitudes that explain voting the impact of social norms, intentions, and past behaviour methodological issues: measurement and research design Implications for the study of voting 59 x

11 PART II: A PSYCHOLOGICAL-PSEPHOLOGICAL SYNTHESIS 4. The Sincere Vote Model 63 Outline of the sincere vote model 65 Concepts of the model and their relationships 66 Use of the model Vote Choice Heuristics 73 Election outcome preference heuristic 74 government preference heuristic government leader preference heuristic party size preference heuristic policy preference heuristic Incumbent approval heuristic 79 Party preference heuristic 81 Candidate preference heuristic 82 Voting habit heuristic 82 Endorsement heuristic 83 The heuristic model of voting Three Models to Explain Party Evaluations 89 The orthodox model 90 a traditional approach to explain party evaluations the orthodox model of party evaluations Kelley and Mirer: the simple act of voting The on-line model 95 the on-line model of candidate evaluations anomalies in the on-line model The emotion-integration model 100 the impact of emotions on candidate evaluations the emotionintegration model of party evaluations xi

12 PART III: ANALYSIS OF FOUR DUTCH PARLIAMENTARY ELECTIONS 7. Empirical Test of the Sincere Vote Model 109 Operationalisation of the concepts 110 measurement of party evaluations measurement of party preferences measurements of voting intentions and voting behaviour A test of the sincere vote model 126 relationships between party preferences, voting intentions, and voting behaviour patterns of relationships the impact of party preference strength Summary and conclusion The Non-Sincere Vote 137 The impact of election outcome preferences 138 The impact of incumbent approval 144 The impact of candidate evaluations 149 The impact of voting habits 157 Multivariate analyses 161 Multiple party preferences: heuristics as tie-breaker 168 Summary and conclusion Explaining Party Evaluations: A Traditional Approach 171 The impact of religious and social class identity 172 The impact of policy preferences 178 The impact of ideology in terms of left-right 183 The impact of government satisfaction 186 The impact of party leader evaluations and multivariate analyses 189 Summary and conclusion 196 xii

13 PART IV: CONCLUSIONS 10. A Psychological Theory of Voting 201 A two-stage model of voting 201 The sincere vote model 206 Explaining party evaluations 209 The psephological paradigm 213 Future research 215 On partisanship 217 APPENDICES A. The concept of a sincere vote 223 B. Illustrations of the use of heuristics 227 C. The conceptualisation of memory 231 D. The conceptualisation of emotions 233 E. The impact of party size 239 F. Party leader evaluations and vote choice 245 G. Implications for psychology 253 NOTES, BIBLIOGRAPHY, AND AUTHOR INDEX Notes 259 Bibliography 289 Author index 305 Nederlandse samenvatting (Dutch summary) 311 Curriculum vitae 329 xiii

14

15 LIST OF FIGURES AND TABLES FIGURES 2.1 The psychological black-box of sociological models of voting Concepts to be distinguished in studies of voting The sincere vote model Sub-types of the election outcome preference heuristic Six heuristics to decide how to vote The heuristic model of voting A traditional model of voting (an example) A traditional model of party evaluations and voting (an example) The orthodox model of voting The orthodox model of party evaluations The on-line evaluation model of voting The emotion-integration model of party evaluations Evaluation of the leader of the preferred party and the chance of a non-sincere voting intention Evaluation of the best-liked leader of non-preferred parties and the chance of a non-sincere voting intention Difference between evaluations of the leaders of preferred and non-preferred parties and the chance of a non-sincere voting intention A two-stage model of voting 205 C.1 Taxonomy of memory 232 D.1 Taxonomy of emotion 236 E.1 Size of the preferred party and the chance of a non-sincere voting intention 243 G.1 Three categories of attitudes and their relationships 254 xv

16 TABLES 7.1 Percentage of voters who knew (how much they liked) the parties Evaluation scores awarded to parties in Evaluation scores awarded to parties in Evaluation scores awarded to parties in Evaluation scores awarded to parties in Percentage of voters who awarded parties evaluation scores of 80 or more Highest evaluation score awarded to any party Preferences for parties in Preferences for parties in Preferences for parties in Preferences for parties in Number of parties in party preferences Distribution of single party preferences Distribution of multiple party preferences Strength of party preferences Distribution of voting intention categories Party preference strength and voting intention categories: percentage of voters who knew for whom to vote Distribution of voting intentions Distribution of voting behaviour Relationship between party preferences and voting intentions voters with single party preferences Relationship between party preferences and voting intentions voters with multiple party preferences Percentage of voters with sincere and non-sincere voting intentions Relationship between voting intentions and voting behaviour voters with single party preferences Relationship between voting intentions and voting behaviour voters with multiple party preferences Percentage of voters who voted as initially intended and who did not Relationship between party preferences and voting behaviour undecided voters with single party preferences Relationship between party preferences and voting behaviour undecided voters with multiple party preferences Percentage of undecided voters who voted sincerely and non-sincerely Relationship between party preferences and voting behaviour voters with single party preferences Relationship between party preferences and voting behaviour voters with multiple party preferences 130 xvi

17 7.31 Percentage of voters who voted sincerely and non-sincerely Patterns of party preference voting intention voting behaviour relationships Party preference strength and non-sincere voting intentions Party preference strength and changes in voting intentions Percentage of voters who preferred specific coalitions Percentage of voters who preferred specific parties in the coalition Percentage of voters who preferred their party preference in the coalition Relationship between party preferences and coalition preferences Coalition preferences and the party preference voting intention relationship Relationship between coalition preferences, party preferences, and voting intentions Parties that participated in government coalitions ( ) Percentage of voters who were satisfied or dissatisfied with the government Relationship between government satisfaction and party preferences (I) Relationship between government satisfaction and party preferences (II) Government satisfaction and the party preference voting intention relationship Relationship between government satisfaction, party preferences, and voting intentions Names and parties of the leaders who were evaluated Relationship between party leader evaluations and party evaluations Evaluation scores awarded to leaders of preferred parties Evaluation scores awarded to leaders of non-preferred parties Differences in evaluations of leaders of preferred and non-preferred parties Relationship between party leader preferences, party preferences, and voting intentions Percentage of voters who recalled their previous vote choice Relationship between previous vote choice and party preferences Previous vote choice and the party preference voting intention relationship Relationship between previous vote choice, party preferences, and voting intentions Which concepts non-sincere voting intentions were in line with A multivariate model of non-sincere voting intentions A multivariate model of non-sincere voting behaviour 165 xvii

18 8.26 Percentage of voters for whom each heuristic could break the tie Percentage of voters who broke tie as expected on the basis of each heuristic Percentage of voters who considered themselves member of a particular church Percentage of Christian church members who attended religious services with a particular frequency Percentage of voters who assigned themselves to a particular social class The impact of religious and social class identity on party evaluations Policies about which voters indicated their positions The impact of policy preferences on party evaluatuions The impact of left-right agreement on party evaluations The impact of government satisfaction on party evaluations The multivariate model and party evaluations Explanatory power of a multivariate model that includes party leader evaluations Percentage of voters who met the expectations based on the sincere vote model and why others did not Explanatory power of various models to explain party evaluations (I) Explanatory power of various models to explain party evaluations (II) 212 B.1 Vote choice motivations related to the election outcome preference heuristic 228 B.2 Vote choice motivations related to the incumbent approval heuristic 228 B.3 Vote choice motivations related to the party preference heuristic 228 B.4 Vote choice motivations related to the candidate preference heuristic 229 B.5 Vote choice motivations related to the voting habit heuristic 229 B.6 Vote choice motivations related to the endorsement heuristic 229 E.1 Size of parties 240 E.2 Party preferences and the party preference voting intention relationship 241 E.3 Voting intentions of voters with a non-sincere voting intention 242 E.4 Relationship between size of preferred party and size of party intended to vote for (voters with non-sincere voting intentions only) 244 xviii

19 F.1 Evaluation scores awarded to party leaders in F.2 Evaluation scores awarded to party leaders in F.3 Evaluation scores awarded to party leaders in F.4 Evaluation scores awarded to party leaders in F.5 Evaluation scores awarded to preferred party leader 250 F.6 Number of leaders preferred 250 F.7 Distribution of single party leader preferences 251 F.8 Distribution of multiple party leader preferences 251 F.9 Distribution of party leader preferences (single or multiple) 252 F.10 Relationship between party leader preferences and voting intentions 252 N.1 Evaluation of the leader of the preferred party and the chance of a non-sincere voting intention 279 N.2 Evaluation of the best-liked leader of non-preferred parties and the chance of a non-sincere voting intention 279 N.3 Difference between evaluations of the leaders of preferred and non-preferredparties and the chance of a non-sincere voting intention 279 N.4 Relationship between voters policy preferences in N.5 Relationship between voters policy preferences in N.6 Relationship between voters policy preferences in N.7 Relationship between voters policy preferences in N.8 Relationship between the size of the preferred party and the chance of a non-sincere voting intention 288 xix

20

21 PART I Introduction and Background

22

23 CHAPTER 1 INTRODUCTION Why do people vote as they do? That is the central question of this research. Why do voters support one candidate or party rather than another? This has been one of the classic questions of political science, which should be no surprise given the fact that elections are the cornerstone of the democratic system. The essence of democracy is that political power is acquired through a competitive struggle for the people s vote. 1 Consequently, without insight in why people vote as they do in elections, we know little about the functioning of democracies. The validity of the interpretation of an election outcome as a mandate, for example, depends on why people vote as they do. If they vote for a party because they approve of its performance in the past, the outcome cannot be interpreted in the same way as when they do so because they like its promises for the future. 2 To understand what meaning can be attributed to election outcomes, and judge how democracies function, we need to understand what makes voters decide in a particular way. The question why people vote as they do is not new. Many answers have already been given and substantial insights have been gained (see Dalton and Wattenberg 1993; Harrop and Miller 1987). Nevertheless, our understanding of voting behaviour is still limited. In a sense, today it appears even more limited than some decades ago. For example, in the Netherlands in the 1950s the choice of a large majority of voters could be explained on the basis of their religiosity and social class (Lipset and Rokkan 1967; Lijphart 1974; Andeweg 1982). Today, however, on the basis of the same information vote choices can be predicted much more poorly. Moreover, no other social characteristics seem to have taken the place of religiosity and social class (Andeweg 1982, 1995; Irwin and Van Holsteyn 1989a; Van Holsteyn and Irwin 2003; Van der Kolk 2000). Alternative explanations of vote choice have since been provided, for example on the basis of ideology and policy preferences (Van der Eijk and Niemöller 1983; Irwin and Van Holsteyn 1989b; Middendorp 1991; Van 3

24 4 THE SINCERE VOTE Wijnen 2001). However, this has not yet resulted in satisfactory explanations. Moreover, since the mid-1980s the explanatory power of models based on ideology and policy preferences seems also to have decreased (Thomassen et al. 2000; Van Wijnen 2001; Van Holsteyn and Irwin 2003). The decrease in understanding described above is the result of changes in society, not developments in scientific theory. One may argue that to know that today religiosity and social class structure vote choice to a limited extent and to know that in the 1950s they did to a large extent, both concern the same amount of understanding. What the example illustrates, is that a so-called sociological approach to voting behaviour is of limited value today. We must turn to other approaches. One possibility is to focus more strongly on what goes on in voters minds. This shift is also important for another reason. The question how well a model explains voting behaviour is often answered on the basis of the number of voters whose choice can be predicted correctly on the basis of information other than their voting in statistical terms: the amount of variance explained of the dependent variable (vote choice) on the basis of the independent variables (for example, religiosity and social class). At least equally important, however, is to understand how independent variables are related to vote choice (Asch 1952, ch. 18). For example, if Catholic voters were all to vote for a Catholic party, the question remains why they do so. Do they vote out of habit, or do they deliberately make up their mind? Do they vote because people in their surrounding influence them, or because they agree with the policy proposals of the party? Or do they vote because they reckon that they share the same ideology? To really understand why people vote as they do, such questions must be answered. The questions just posed point to one specific way in which our understanding of voting is limited, namely in terms of the underlying psychological processes. Although much electoral research has been done, the psychology of voting has remained poorly understood. Some scholars have explicitly stated that their studies adopted a psychological perspective, most notably those of the so-called Michigan school (Campbell et al. 1954, 1960). In the following chapter it will be argued, however, that it may be doubted whether these studies are as psychological as generally thought, and whether they give insight in the psychological processes that underlie voting. THE RESEARCH QUESTION Voting is often regarded as comprising two decisions: whether to vote or abstain, and which party or candidate to vote for. The two decisions correspond with different questions that voters ask themselves: Shall I vote or not? and For whom shall I vote?. These questions may be related. If voters know whom they prefer, this may

25 Introduction 5 be an incentive to cast a vote. If voters have no idea for whom to vote, this may be a reason to abstain. This implies that the two decisions are not always made independently of one another. Nevertheless, in voting research both decisions are often regarded as independent, or at least treated as if they are (see, for example, Dalton and Wattenberg 1993; Miller and Shanks 1996). This research adopts the common approach and focuses on for whom people vote. Hence, in this study the question is why people vote as they do, given the fact that they vote. If voting behaviour is defined in terms of for whom people vote, the question arises who this whom is. In this research voting is defined in terms of parties. While in principle the theory to be outlined can also be applied to other objects for example, candidates this study speaks about vote choice in terms of parties for whom people vote. The main reason to do so is that in most established democracies parties play a key role. Elections may therefore be conceived of as competitions between parties. Moreover, even if, in a formal and/or psychological sense, people vote for candidates, their votes may still be analysed in terms of the parties that in a sense receive them. This means that the research question may in practice be phrased as why people vote for (a candidate of) a particular party. A PSYCHOLOGICAL APPROACH The fact that the study of voting has its own name psephology 3 illustrates that one might regard the study of voting as a discipline on its own. The theories that have been used to explain voting are usually not meant to be applied outside the voting domain. This does not mean, however, that psephology operates in isolation of other fields. Rather, the major theories of voting have a basis in three other disciplines: sociology, psychology, and economics (Harrop and Miller 1987). Likewise, this study is strongly based in psychology. The psychological explanation for voting fits the fourth and latest phase of the development of the study of mind and behaviour (see LeDoux 1998, ch. 2). Until the beginning of the twentieth century the study of mind and behaviour was based primarily on introspection. During the first half of the twentieth century the dominant approach became that of behaviourism, which proclaimed that concepts that concern the mind, and which were gained access to by introspection (e.g. concepts like perception, memory, and emotion), are not appropriate topics for scientific study. In the behaviourist approach the psychological processes that mediate between external events and behaviour were regarded as a black box. With the development of computers a new metaphor for the human mind arose, namely that of an information-processing system. This laid the base for a third approach, which opened up the black box: cognitive science. An essential element, however, was still missing. Arguably, the most important difference between computers and humans is that, unlike

26 6 THE SINCERE VOTE humans, computers are not directed by emotions. The incorporation of emotions in information-processing marks the approach of the fourth phase. Unlike the behaviourist approach, the study of mental phenomena is regarded as necessary; and unlike the purely cognitive approach, emotions are regarded as phenomena to be included. The core idea in the psychological approach adopted is that to understand why people vote as they do, we must understand what goes on in their mind. When individuals act, like when they vote, specific thoughts or feelings have preceded and these are regarded as the key to understanding the behaviour. This is not to say that the underlying psychological processes are all there is to know. However, even for the understanding of the impact of non-psychological phenomena, such as social settings and political events, insight in the psychological processes is important. Such phenomena have an impact on voting behaviour only if (and thus because) they influence how voters think and feel about political objects. This means that their impact is mediated by psychological variables. Hence, the study of social settings and political events may not only benefit from, but arguably requires, insight in the psychological processes that underlie voting. 4 Psychological variables may be regarded as consequences of non-psychological variables and as causes of behaviour. 5 However, psychological variables may also be consequences of behaviour. Various scholars have emphasised that what individuals do influences what they feel or think, rather than the other way round (Festinger 1957; Bem 1972). In this study those processes will not be focused on. The question here is what psychological processes underlie voting behaviour, not what psychological processes result from it. However, when focusing on particular psychological variables in order to explain voting, it is important to ask the question whether these variables should not be regarded consequences, rather than causes, of the behaviour studied. Furthermore, it is important to note that causal relationships may also exist between psychological variables themselves. 6 This implies that psychological variables may be related to behaviour indirectly, since their influence may be mediated by other psychological variables. So to understand the psychological processes that underlie voting, the way psychological variables are related to one another must also be examined. The question which psychological variables are to be focused on remains to be answered. This will be done in the remainder of this study by applying some general ideas from social psychology, and to some extent neuroscience, to the electoral context. Theories not directed at voting behaviour may provide useful insights for its study. In some instances the electoral context may require modifications of those theories. Another difference with general psychological research is that in this study the concepts will mostly be phrased in a way that only applies to voting. For example, a key concept in this study is that of party evaluations. These may be seen as an example of the social psychological concept of attitudes.

27 Introduction 7 THE DUTCH CASE The psychological processes that underlie voting are presumably not fundamentally different across voters, elections, countries, or time. All voters use similar brains and therefore similar psychological processes will operate. Consequently, in a sense it is not relevant in which context the theory to be developed is put to the test. Nevertheless, from the perspective of electoral research, the Netherlands is particularly interesting. As will be shown in more detail in the next chapter, the Netherlands has been considered a country to which a sociological approach applies very well (Lipset and Rokkan 1967: 15). In terms of so-called psychological models of voting the Dutch case was found to be special too, because the models could not be applied well. With respect to party identification, one of the central concepts in psychological models of voting, the Netherlands has been found to be the oddest case of all (Miller and Shanks 1996: 117; see Chapter 2). Among Dutch voters party identification and electoral choice could not be distinguished meaningfully (Thomassen 1976b), and the major psychological model was therefore considered not useful. Although solutions to the Dutch party identification problem have been proposed, as yet it has not been solved. To substantiate the principal claim of this study that vote choice can be distinguished meaningfully from partisanship if it is conceptualised in terms of party evaluations, a useful strategy may be to test the ideas in a context that seems least likely to be suitable. Because the Dutch case has been known for its unsuitability to explain voting behaviour on the basis of the existing psychological models and the impossibility to distinguish between partisanship and vote choice, there is arguably no better case for an empirical test than the Dutch one. Dutch parliamentary elections may also be considered interesting in their own right. Although Dutch politics has perhaps not been renowned for a high level of excitement, recent elections brought some eye-catching changes. In 1994 the Christian Democrats (CDA) suffered a record loss of 20 seats (out of 54). This resulted in the first government without any Christian party since the introduction of universal suffrage at the beginning of the twentieth century. 7 In 2002 the Labour Party (PvdA) beat the record, losing 22 of their 45 seats. The most amazing electoral event was arguably the fact that in the same year a new list of candidates headed by Pim Fortuyn, who was assassinated nine days before the election, entered parliament with 26 seats. This made them the second-largest party. To understand these events, it is necessary to gain insight in what went on in the minds of voters. A final reason why the Dutch case is an appropriate one for this study, is that data for analysis are available. The Netherlands has a research project called the Dutch Parliamentary Election Studies (DPES), which contains data from surveys held around each parliamentary election since 1971 (about the DPES, see Van der Eijk and Niemöller 1994). On the basis of these data, the major models of voting have

28 8 THE SINCERE VOTE already been tested. Although the DPES data are not perfect for testing all ideas put forward in this research, the surveys in 1986, 1994, 1998, and 2002 included questions on the basis of which key ideas can be tested. DESIGN OF THE BOOK The aim of this research is to contribute to the understanding of voting at the theoretical level by focusing on the psychological processes that underlie voting. A theory of voting will be developed and the corresponding models will be tested empirically in the context of Dutch parliamentary elections. This book is organised in four parts: background, theory, empirical analyses, and conclusions. The psephological and psychological background of this research will be discussed in the following two chapters. Chapter 2 points out how existing theories of voting might be viewed from a psychological perspective, what their implications are, and where they go wrong. Chapter 3 discusses attitude-behaviour research, an important sub-field of social psychology, including applications to voting behaviour. The aim of Part II is to synthesise insights from voting research and psychological research in order to develop a psychological theory of voting. Chapter 4 deals with the relationship between party evaluations and voting behaviour, which are specified in the sincere vote model. Chapter 5 discusses the choice mechanisms that underlie voting in terms of six ways in which voters may decide how to vote. Chapter 6 concerns the explanation of party evaluations. In Part III the theory is put to the test in the context of four Dutch parliamentary elections. Chapter 7 presents an empirical test of the sincere vote model. Chapter 8 examines why some voters cast a so-called non-sincere vote by focusing on alternative choice mechanisms. Chapter 9 contains analyses of why voters evaluated parties as they did. The concluding Chapter 10 in Part IV integrates the theoretical models, summarises the findings of the empirical analyses, and discusses their implications.

29 CHAPTER 2 PSYCHOLOGY IN VOTING THEORY Three approaches to voting are distinguished: a sociological, psychological, and economic approach (Harrop and Miller 1987, ch. 6). These are represented by the Columbia school, the Michigan school, and the Downsian school, respectively. This chapter elaborates upon the corresponding studies. This discussion is not an extensive review of all theories of voting, but a highlighting of the major approaches (for more thorough reviews, see Dalton and Wattenberg 1993; Harrop and Miller 1987). Discussing the so-called Michigan studies is especially relevant, since these are regarded as the most important example of a psychological approach. Special attention will also be given to theories of issue voting and to applications to the Netherlands. The discussion in this chapter focuses on the psychology in the theories. Both the explicit and the implicit assumptions and hypotheses about the psychology of voting are elaborated upon. This should further clarify what the psychological perspective that is adopted in this research means. What is important in this respect, is that some theories that are often regarded as psychological will be shown to be in a sense non-psychological, while some other theories that are regarded as non-psychological will be shown to be in a sense psychological. ON ELECTORAL RESEARCH DEVELOPMENT OF MODERN ELECTORAL RESEARCH Today, electoral research is based primarily on large-scale projects held when major elections take place. In the United States large-scale surveys are held around the presidential elections: the American National Election Studies. In many other countries there are similar projects, such as the British Election Studies and the Dutch 9

30 10 THE SINCERE VOTE Parliamentary Election Studies. These studies have one feature in common: they are based on interviews with a large sample of voters. Individual data have not always been the standard for election studies. Before surveys were the basis of analysis, electoral behaviour was examined mainly on the basis of census data or comparable sources of information. Not the individual, but geographical entities were the unit of analysis. This changed with the work by the commercial polling agencies, which in the 1930s collected individual data about electoral behaviour on a large scale for the first time (Campbell et al. 1960: 14). Their work laid the foundation for major voting studies in the United States. The first election to be studied extensively in this way was the 1940 U.S. presidential election. Paul Lazarsfeld, Bernard Berelson and Hazel Gaudet from Columbia University published the results in 1944 in The People s Choice. This study marks the beginning of modern electoral research. At the theoretical level the unit of analysis shifted to the individual, replacing geographical entities such as regions or neighbourhoods. At the methodological level it marks the beginning of the use of interviewing and the survey method. The aim of The People s Choice was to examine the psychological process of opinion formation, but the authors ultimately concluded that vote choice could be explained well on the basis of social characteristics. This sociological research left many questions unanswered in particular, how the voter s mind was affected and from a psychological perspective it did not answer the question why people vote as they do satisfactorily. This partly changed in 1954, when Bernard Berelson, Paul Lazarsfeld and William McPhee published a second Columbia study, Voting. In this study it was explained why social characteristics influenced vote choices. Two processes or mechanisms that linked social characteristics to vote choices were considered specifically important: social influence on voters by people who surround them, and political interests shared by voters with similar social characteristics. In the same year, Angus Campbell, Gerald Gurin and Warren Miller from the Survey Research Center of the University of Michigan published their first major election study, The Voter Decides. The Michigan scholars argued that their research shifted the attention from sociological to psychological variables. They focused on voters stands on the issues, their feelings about the major parties, and their perceptions of the candidates. This approach was elaborated upon by Angus Campbell, Philip Converse, Warren Miller and Donald Stokes in The American Voter, published in The Michigan studies have since been considered the major representatives of a psychological approach to voting behaviour (Harrop and Miller 1987, ch. 6). An economic approach is usually distinguished as a third approach. An Economic Theory of Democracy, published in 1957 by Anthony Downs, is regarded its major representative. Downs discussed how voters would behave if they applied the principles of rationality, which he derived from economic theory, to voting. Various scholars have tested Downs ideas empirically. Like the so-called sociological and

31 Psychology in Voting Theory 11 psychological studies, these studies were based on survey data and had the individual as unit of analysis. THE FORCE FIELD FRAMEWORK Theories of voting are usually phrased in terms of one of two analytical frameworks: the force field framework and the spatial framework. The force field framework is related to the work of the psychologist Kurt Lewin. According to Lewin s (1951) field theory, behaviour results from a number of simultaneous psychological forces that act upon a person. The net effect of these forces determines how individuals behave. With respect to voting this implies that voters experience various forces that direct them either towards voting for a particular party, or away from such voting. Together such forces determine for whom they vote. In The People s Choice the Columbia scholars adopted the force field framework. Their discussion of what they called cross-pressures clearly shows this. By cross-pressures we mean the conflicts and inconsistencies among the factors which influence vote decision. Some of these factors in the environment of the voter may influence him toward the Republicans while others may operate in favor of the Democrats. In other words, cross-pressures upon the voter drive him in opposite directions. (...) There were a number of factors differentiating Republican and Democratic voters. Each of these factors could be considered a pressure upon final vote decision. We found the Protestant vote allied to the Republicans and the Catholic vote more strongly Democratic. We found that individuals on the higher SES levels tended to vote Republican and their poorer neighbors to vote Democratic. In other words, a vote decision can be considered the net effect of a variety of pressures. (Lazarsfeld et al. 1944: 53, 56) The notions of forces and pressures are fairly similar. The words quoted also indicate what the pressures stemmed from: social characteristics such as religious affiliation and socio-economic status. Forces related to these characteristics predisposed voters in a Democratic or Republican direction. Although cross-pressures were predominantly viewed as resulting from social characteristics, the Columbia scholars used the term also to discuss attitudinal forces. Some voters (...) were subject to strong attitudinal cross-pressures (resulting from their weak predispositions); they liked Roosevelt for this and Willkie for that, or they approved one part of a candidate s program but disapproved another part. (Lazarsfeld et al. 1944: 99) So the notion of forces was in some instances related to voters psychological, rather than social, characteristics.

32 12 THE SINCERE VOTE In The Voter Decides the Michigan scholars spoke about motivational forces and psychological forces, indicating the use of the force field framework. Party identification, issue orientation, and candidate orientation were regarded as three simultaneous forces acting upon individuals and influencing their vote choice. In The American Voter the Michigan scholars explicitly linked their theory to Lewin s field theory (Campbell et al. 1960: 33). Forces were conceptualised in terms of six so-called partisan attitudes, each of which could be regarded as a force towards either a Republican or a Democratic vote. Both the Columbia and Michigan scholars did not assume that voters experienced the forces they distinguished as such. The forces were not psychological entities, but analytical constructs. In Voting the Columbia scholars stated that situations with cross-pressures (...) may have no subjective reality for the voter himself at the moment; it is merely that we as observers can foresee his future vulnerability by virtue of his location within cross-currents of social influence. (Berelson et al. 1954/1966: 284) In The Voter Decides the Michigan authors stated that their three-fold division of orientations was also made for analytical purposes. In defining the concept of candidate orientation, as in our definitions of party identification and issue orientation, we are attempting to differentiate analytical constructs, and are not attempting to isolate psychologically pure types. (Campbell et al. 1954/1971: ) This view can be extended to The American Voter. The partisan attitudes were not psychological entities either, but analytical constructs to examine the influence of voters perceptions and feelings with respect to candidates, parties, and issues. THE SPATIAL FRAMEWORK Another framework that has been used in voting research is that of spatial analysis. The central idea here is that parties and voters can both be positioned in some kind of (political) space in which mutual distances represent differences and similarities: the larger the distance between two objects, the larger their difference; the smaller the distance, the larger their similarity. The dimensions of the space usually represent certain dimensions of political conflict. If politics is viewed as being structured by only one dimension of conflict, we speak about a one-dimensional space. In a similar way two dimensions of conflict correspond with a two-dimensional space, three with a three-dimensional space, and so on. Voting models using this framework are usually based on the assumption that voters choose the party closest to them in the political space. This idea is known as the smallest distance hypothesis.

33 Psychology in Voting Theory 13 In An Economic Theory of Democracy (1957) Downs made use of the spatial framework and defined it in terms of ideology. A simple conceptualisation would be to view the political space as one-dimensional based on ideology in terms of a left-right continuum. Parties and voters are then assumed to take a position on this dimension, and voters are assumed to vote for the party that is closest to them. In addition to ideology, other phenomena may be used to define the political space; for example, a number of salient issues. The number of issues involved would then correspond to the number of dimensions of the space. Issues and ideology can both be regarded as dimensions of political conflict. 1 The spatial framework may also be used to model the impact of other phenomena, such as voters images of candidates personal capabilities (Enelow and Hinich 1984, ch. 5; Endersby 1994). Studies of voting that make use of the spatial framework are mostly silent about the psychological processes that operate. Their aim is not to reveal what goes on in voters minds, but to explain voters choices at the polls on the basis of their ideological positions or policy preferences. Through which psychological processes the latter are transformed into a vote choice, is not considered relevant. The spatial framework can be used to explain voting behaviour, but it is not a theory of voting. The question how the political space is to be defined is not answered by merely adopting the spatial framework. Similarly, the question what forces constitute a field is not answered by adopting the force field framework. Only after the dimensions of the space, or the forces of the field, are defined, is an answer given to the question why people vote as they do. The most relevant question when discussing theories of voting is therefore not which of the two analytical frameworks is used, but what the forces or distances correspond with or stem from. If one defines forces in relation to social characteristics, then social characteristics are the explanation provided for voting. If forces are defined in terms of attitudes, then attitudes are the explanation. If one focuses on a one-dimensional ideological space, then ideology is the explanation provided for voting. If dimensions are defined in terms of policy preferences, then these constitute the explanation. What is also relevant, is which social characteristics are put forward, which attitudes, which ideological dimension, and which policies. The remainder of this chapter will focus on these aspects and additionally will elaborate further upon what is psychological, and what is not, about the various studies. A SOCIOLOGICAL PERSPECTIVE: THE COLUMBIA STUDIES SOCIAL CHARACTERISTICS AND POLITICAL PREDISPOSITION The People s Choice (1944) was based on repeated interviewing of the same people during the campaign period. Seven interviews were held with a sample of voters

34 14 THE SINCERE VOTE from Erie County, Ohio, in the period from May until November 1940, when Franklin D. Roosevelt eventually beat his Republican challenger Willkie to secure his third term as U.S. President. The aim of the study by Lazarsfeld and his colleagues was to discover the processes underlying opinion formation and political behavior (Lazarsfeld et al. 1944: 10), which is reflected in the subtitle of the study: How the Voter Makes Up His Mind in a Presidential Campaign. The study concentrated on voters exposure to the campaign in the media, and on changes in voting intentions and the reasons underlying these. It made extensive use of open-ended questions about why voters changed their vote choice and why they favoured a particular candidate. The best known findings of the study, however, stem from a chapter that examined the impact of various social characteristics, among which socio-economic status, religious affiliation, and residence. The general pattern was that voters with a high socio-economic status favoured the Republicans, those low in socio-economic status the Democrats; Protestants favoured the Republicans, Catholics the Democrats; rural voters favoured the Republicans, urban voters the Democrats. On the basis of these social characteristics a so-called index of political predisposition was constructed. Voters with a high SES level, who were Protestant, and who lived in the rural area had the strongest Republican predisposition, whereas urban Catholics with a low SES level had the strongest Democratic predisposition. A clear empirical relationship between political predisposition and voting was found: when voters predisposition tended stronger towards the Republican or Democratic side, they voted for the corresponding party in larger numbers. Whether they would vote for the Republican or Democratic candidate could be predicted correctly on the basis of the index for 67 per cent of the voters. 2 The main conclusion of the study consequently was that voters choices were largely determined by their social characteristics. The authors came to their much-cited conclusion that a person thinks, politically, as he is, socially. Social characteristics determine political preference (p. 27). The People s Choice showed that, and which, social characteristics were related to vote choice. The second Columbia elaborated upon why they were. In Berelson, Lazarsfeld and McPhee s Voting (1954) the central idea was that people are influenced by whom they have contact with, which in turn is determined by their social characteristics. Three principles related to this process were identified: (1) there is a social basis for political interest: voters belong to different groups that have different interests, (2) voters inherit vote preferences from their family, and (3) voters have contact predominantly with other people from the same social groups (cf. pp , 147). By these three principles the groups people belong to influence for whom they vote. 3 Hence, because they have similar political interests a political predisposition based on social characteristics translates into vote preferences. 4 Another part of Voting focused on issues, since these supplied the content of political debate [and were] the stuff in terms of which a democratic campaign is rationalized, in both senses (p. 182). The authors also discussed effects of candidate